Browse Prior Art Database

System and Method for Interactively Identifying Regions of a Digital Image

IP.com Disclosure Number: IPCOM000117860D
Original Publication Date: 1996-Jun-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 2 page(s) / 96K

Publishing Venue

IBM

Related People

Barber, R: AUTHOR [+3]

Abstract

Many image processing functions require that regions or objects in an image be identified or outlined. The outlining can be done by manual, automatic, or semi-automatic methods. Flood fill is one semi-automatic method. However, existing flood-fill methods often capture too much or too little of an object, have adjustable parameters that can be difficult to set, and have awkward facilities to correct an outline once it is made. This disclosure describes an interactive, easily-controlled flood-fill methodology.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

System and Method for Interactively Identifying Regions of a Digital
Image

      Many image processing functions require that regions or objects
in an image be identified or outlined.  The outlining can be done by
manual, automatic, or semi-automatic methods.  Flood fill is one
semi-automatic method.  However, existing flood-fill methods often
capture too much or too little of an object, have adjustable
parameters that can be difficult to set, and have awkward facilities
to correct an outline once it is made.  This disclosure describes an
interactive, easily-controlled flood-fill methodology.

      Image processing applications frequently require that objects
be identified or outlined in a displayed image.  Fully automatic
outlining is unreliable so many systems use manual or semi-automatic
methods.  One such method is called flood-fill.  In this method, the
user selects a pixel (or area) as an object pixel.  The following
process is then performed repeatedly until no additional pixels are
added: any pixel touching an object pixel and whose value is within
some range (e.g., plus or minus 10) of the value of the original
object pixel is added to the set of object pixels.  Flood-fill
techniques typically have the following problems:
  1.  They require user tunable parameters controlling the range.
       There may be one range as described above or, for color
images,
       one range for each color channel, and all need to be adjusted.
       Adjusting these ranges is often tedious.  The proper setting
is
       dependent on the image material surrounding the object, the
       uniformity and contrast between the object and background,
       shading, texture, and other factors.  Typically the value(s)
need
       to be adjusted separately for each identified object, and user
       input is required to set, change, undo, and retry the filling
       with new setting, and the process is repeated for each new
       object.
  2.  "Leakage" can occur, in which adjacent objects of similar color
       get included in the fill.  This often happens because there is
       some small set of pixels that meet the range criteria
connecting
       the two objects, and the fill algorithms "leaks" through in
these
       pixels.  Image noise, shading, and specular reflections often
       produce pixel values that allow leakage.  This is particularly
a
       problem when an object has low-contrast edges.
  3.  "Holes" can occur in the filled object, which are subareas of
the
       object of a different color.  This is common in textured
objects
       object of a different color.  This is common in textured
objects
       that contain a range of different grey level or color values.
       These differences can result in an object with many s...